When writing programs to automate – well, everything – always use shell scripts. Because shell scripts are stored in ASCII text format, they can be easily viewed and modified by humans just as easily as they can by computers. You can examine a shell program and see exactly what it does and whether there are any obvious errors in the syntax or logic. This is a powerful example of what it means to be open.
I know some developers tend to consider shell scripts something less than true programming. This marginalization of shell scripts and those who write them seems to be predicated on the idea that the only true programming language is one that must be compiled from source code to produce executable code. I can tell you from experience this is categorically untrue.
I have used many languages including BASIC, C, C++, Pascal, Perl, Tcl/Expect, REXX, and some of its variations including Object REXX; many shell languages including Korn, csh, and Bash; and even some assembly language. Every computer language ever devised has had one purpose – to allow humans to tell computers what to do. When you write a program, regardless of the language you choose, you are giving the computer instructions to perform specific tasks in a specific sequence.
Scripts can be written and tested far more quickly than compiled languages. Programs usually must be written quickly to meet time constraints imposed by circumstances or having to rush to get ready for your UK vacation. Most of the scripts we write are to fix a problem, to clean up the aftermath of a problem, or to deliver a program that must be operational long before a compiled program could be written and tested.
Writing a program quickly requires shell programming because it allows quick response to the needs of the customer whether that be ourselves or someone else. If there are problems with the logic or bugs in the code, they can be corrected and retested almost immediately. If the original set of requirements was flawed or incomplete, shell scripts can be altered very quickly to meet the new requirements. So, in general, we can say that the need for speed of development in the SysAdmin’s job overrides the need to make the program run as fast as possible or to use as little as possible in the way of system resources like RAM.
Most things we do as SysAdmins take longer to figure out how to do than they do to execute. Thus, it might seem counter-productive to create shell scripts for everything we do. Writing the scripts and making them into tools that produce reproducible results and which can be used as many times as necessary takes some time. The time savings come every time we can run the script without having to figure out again how to perform the task.
You may encounter a situation when the execution of a script takes an excessive amount of time or when the problem is so general that it will be done thousands or even millions of times, in which case compiled languages make more sense. But those are extremely rare cases.